A Novel Orexin Antagonist from a Natural Plant Was Discovered Using Zebrafish Behavioural Analysis

Home , Antagonist, Orexin, Orexin antagonist, Suvorexant

European Review for Medical and Pharmacological Sciences 2020; 24: 5127-5139 A novel orexin antagonist from a natural plant was discovered using zebrafish behavioural analysis

M. YAMANAKA1, H. IWATA2, K. MASUDA2, M. ARAKI2, Y. OKUNO2, M. OKAMURA1, J. KOIWA1, T. TANAKA1

1Department of Systems Pharmacology, Mie University, Graduate School of Medicine, Tsu, Mie, Japan 2Graduate School of Medicine, Kyoto University, Kyoto, Japan

Abstract. – OBJECTIVE: Phenotypic screening methods. However, these methods do not focus ing is one of the most practical approaches on the interactions of the whole organism’s activity, to the identification of mediators of behaviour, resulting in difficulty to ensure success at an organ- since it is difficult to model brain function in vi- ismal level. To overcome these difficulties and to tro, at a cellular level. We used a zebrafish (Da- nio rerio) behavioural assay to discover novel, increase the efficiency of screening, pharmacolog- natural, neuroactive compounds. ical researchers have adopted a phenotype-based MATERIALS AND METHODS: A zebrafish screening method. Many first-in-class drugs have behavioural assay was performed for seven been discovered by the phenotype-based approach, natural compounds, obtained from plants. The rather than the target-based approach1. behavioural profiles were compared to those of Phenotypic screening is one of the most prac- known psychoactive drugs. We characterised a natural compound exhibiting a behaviour profile tical approaches to the identification of mediators similar to that of suvorexant, using in silico, in of behaviour, since it is difficult to model brain vitro and microarray expression analysis. function in vitro, at a cellular level. During the last RESULTS: The behavioural analysis performed decade, zebrafish behaviours have become one of in this study classified central nervous system the major screening systems used to identify neu- drugs according to their mechanism. Zebrafish roactive compounds, through behaviour-based b treated with a natural compound, 8 -(4'-Hydroxy- chemical screening2. Zebrafish are vertebrates, tigloyloxy) costunolide (8b), showed behaviour and the structure and function of their organs profiles similar to those of zebrafish treated with 3,4 suvorexant, a known orexin antagonist. This be- and tissues are similar to those of humans . The havioural assay was validated using in silico and receptors and neurotransmitters are highly con- in vitro assays, which revealed that the new com- served between humans and zebrafish5. pound was a dual orexin receptor antagonist. In In this study, we attempted to discover novel, addition, transcriptome analysis suggested that natural, neuroactive compounds by phenotypic 8b might regulate the nuclear factor-κB (NF-κB) related pathway. screening, using zebrafish behavioural analysis. CONCLUSIONS: We conclude that zebrafish As a test case, we focused on a natural compound phenotypic screening, combined with in silico exhibiting a behaviour profile similar to that of su- assays and gene expression profiling, is a useful vorexant, a well-known orexin receptor antagonist. strategy to discover and characterize novel ther- Orexins are neuropeptides involved in the reg- apeutic compounds, including natural products. ulation of the sleep-wake cycle, feeding pattern, 6,7 Key Words: energy balance, and stress in mammals . Dys- Zebrafish, Behaviour, Phenotype-based screening, regulation or loss of orexin signalling has been Sleep, Orexin. linked to narcolepsy8-10. Several dual orexin receptor antagonists have been investigated as potential treatments for insomnia, and one of them, suvorexant, received approval by the US Food Introduction and Drug Administration (FDA), in 2014, for the treatment of such a condition11. Although these Many successful therapeutics have been dis- drugs have an outstanding safety profile, there are covered using cell-based and biochemical screen- concerns regarding the use of these drugs, due

Corresponding Author: Toshio Tanaka, MD; e-mail: [email protected] 5127 M. Yamanaka, H. Iwata, K. Masuda, M. Araki, Y. Okuno, M. Okamura, J. Koiwa, T. Tanaka to potential side-effects and lack of evaluation of compounds were added to each 48-well plates. long-term effects. Therefore, natural compounds The plates were then placed in the Daniovision could be promising agents, because of their multi- system (Noldus, Wageningen, The Netherlands), functional health effects and limited toxicity. wherein they were blocked from daylight and il- In this study we characterised a natural com- luminated from below with white light (255 lx) pound exhibiting a suvorexant-like profile, based from 4 pm to 7 pm and from 5 am to 4 pm. The on zebrafish behavioural analysis, and deter- behaviour of zebrafish in each well was monitored mined whether phenotype-based screening using using the Daniovision system with a resolution of zebrafish could serve as a screening model for 1280 × 960 pixels at 25 frames per seconds. Eight natural compounds with the desired mechanisms larvae were assigned to examine the effect of each of action. concentration of the compound. Two independent experiments were performed for each compound. The recorded videos and images were subject- Materials and Methods ed to Ethovision XT11 (Noldus) to analyse the behaviour of zebrafish in each well. The mean Ethical Approval velocity, total distance moved, distance to centre The investigation conformed to the Ethical zone (2 mm radius circle) of the well, frequency Guidelines established by the Institutional Animal entering the centre zone, turn angle, and mobil- Care and Use Committee of Mie University, Japan. ity were measured. The parameters used in this behavioural analysis are shown in Figure 1-a. Compounds Mobility was calculated by taking every pixel Haloperidol, Chlorpromazine hydrochloride, identified as the subject and comparing it with the and Imipramine hydrochloride were purchased current image and the previous one. If all the pix- from Tokyo chemical industry (TCI) (Tokyo, Ja- els were the same, this would indicate zero mo- pan). Diazepam and Clonazepam were purchased bility. If all the pixels were different, this would from Sigma-Aldrich (St. Louis, MO, USA). Su- indicate 100% mobility. In this study, we defined vorexant was purchased from Toronto research 5%-35%, 35%-65%, and 65%-95% as low, medi- chemicals (North York, ON, Canada). These um, and high mobility, respectively14. compounds were dissolved in dimethyl sulfoxide Behavioural endpoints were measured: the (DMSO; Nacalai Tesque, Kyoto, Japan) to make distance moved total (DMT), distance to the zone stock solutions. in the centre of the well (DTZ), in zone frequency (IZF), and turn angle (TA) in each mobility Zebrafish Husbandry category for each period (L1-time, the first light In this study, the nacre/rose/fli1: egfp zebrafish time period for 3 h; D-time, dark time period for was obtained by cross-breeding nacre (−/−)/rose 10 h; L2-time, second light time period for 11 h), (−/−) mutants and Tg (fli1: egfp) transgenic zebraf- resulting in 36 measured endpoints. Overview ish as described previously12. The zebrafish were of the behavioural analysis used in this study is incubated at 28°C under a 14 h light: 10 h dark cy- shown in Figure 1-b. The means obtained were cle and, in environmental quality water, according compared by analysis of variance using Prism 7 to The Zebrafish Book13. Embryos were obtained (GraphPad, La Jolla, CA, USA). Alpha was set at via natural mating and cultured in 0.3× Danieau’s 0.05 and Dunnett’s multiple comparisons test was solution [19.3 mM NaCl, 0.23 mM KCl, 0.13 mM used for post hoc analyses when significant dif-

MgSO4, 0.2 mM Ca (NO3)2, 1.7 mM HEPES, pH ferences were observed. 7.2] until 8 days post-fertilization (dpf). For hierarchical clustering, the data for each behavioural endpoint were normalized according Behavioural Analysis to the controls. Hierarchical clustering was per- The behavioural assay was performed on 48 formed using MultiExperiment viewer. zebrafish at 8 dpf. Each zebrafish was placed into a well on a round 48-well plate (10 mm diameter, Natural Product Library for in-silico Assay 300 μl of 0.3× Danieau’s solution) at 12 pm. The To perform an in silico assay, a natural prod- 48-well plates were then placed in an incubator uct library was constructed, consisting of 9334 set at 28°C with constant light (255 lx) from 12 plants-derived compounds, including 8b. The pm to 3:30 pm. After incubation, 300 μl of 0.3× products were collected from AnalytiCon (Pots- Danieau’s solution with or without the sample dam, Germany), App Tec (Shanghai, China),

5128 A novel orexin antagonist from a natural plant was discovered using zebrafish behavioural analysis

Figure 1. Larval zebrafish behavioural assay was scored based on the following parameters: A, Illus- tration of the behavioural analysis parameters; B, Overview of the behavioural analysis used in this study.

BioBioPha (Yunnan, China), Biopurify (Sichuan, consist of suvorexant, almorexant, SB-649868, China), Chemfaces (Hubei, China), and Pharmeks TCS-OX2-2921, and MK-369722. Suvorexant, alm- (Moscow, Russia). orexant, and SB-649868 exhibit antagonist activities towards both OX1 and OX2 receptors. Collection of Known Antagonists for OX1 and OX2 Receptors In silico Assay Using Pharmacophore- We collected known antagonists of OX1 and Search and Docking Simulation OX2 receptors (Supplementary Figure 1). OX1 Pharmacophores of OX1 antagonists were con- antagonists consist of suvorexant15, almorexant16, structed and pharmacophore searches were per- SB-64986817, SB-33486718, SB-408124 and SB- formed using the natural product library in the 67404219, and GSK-105986520. OX2 antagonists Ligandscout software23,24. After 3D conformations

5129 M. Yamanaka, H. Iwata, K. Masuda, M. Araki, Y. Okuno, M. Okamura, J. Koiwa, T. Tanaka of the antagonists were generated, clusters were proximately 1 h at 23°C. To evaluate the antagonist created and categorized into two classes: (a) suv- effect, solutions containing 50 μl of orexin A were orexant, almorexant, SB-649868, and SB-674042, added to the cultured cells in the 96-well plates to (b) SB-334867 and SB-408124. A pharmacophore obtain final culture medium volume of 200 μl. Fi- model was constructed for each class (Supple- nal concentration of orexin A in the culture medi- mentary Figure 1-a, 1-b). Finally, we performed um was 50 nM. After 24 h, 5 μl of cell supernatant a pharmacophore search from the natural product were transferred from each well to a fresh 96-well library using both the pharmacophore models. plate. After incubation at 65°C for 30 min, 50 μl of The same methods were employed to search substrate were added to each well; the wells were for OX2 antagonists. The known OX2 antagonists mixed and subsequently incubated at 23°C for 30 were categorized into two groups: (c) suvorexant min. The luminescence intensity of the secreted and SB-649868, (d) almorexant and TCS-OX2-29. SEAP (Secreted Alkaline Phosphatase) was mea- A pharmacophore model was constructed for each sured using the SpectraMax i3 platform (Molecu- group (Supplementary Figure 1-c, 1-d), which lar Devices, San Jose, CA, USA). was then followed by the pharmacophore search. Crystal structures of the OX1 (PDBID: 4ZJ8) and Orexin activity rate was calculated using the OX2 (PDBID: 4S0V) receptors forming a com- following formula: plex with suvorexant were obtained from the Activity ratio = (C-Cposi)/ (Cneg-Cposi) Protein Data Bank (PDB) database25-27. After the bound ligand was removed, the structures of the Where C was reported in rectified linear unit disordered loops and flexible side chains were (RLU) in the presence of a test antagonist plus modelled, and the dominant protonation state at orexin A. Cposi was reported in RLU in the pres- pH 7.0 was assigned for titratable residues using ence of suvorexant (as positive control) plus orexin the structure preparation module in the Molec- A, and Cneg was reported in RLU in the presence ular Operating Environment (MOE, Chemical of orexin A only. In this assay, suvorexant was Computing Group, Montreal, Canada), version added at a concentration of 1 μM. Experiments 2013.08. This was followed by molecular dock- were carried out in triplicate. Data were analysed ing of high-scoring compounds obtained from the using the Student’s t-test. A p-value less than 0.05 pharmacophore search. The compound binding was considered statistically significant. site in the OX1 or OX2 receptor was defined to include all atoms within 10 Å from bound suv- RNA Extraction orexant. Each of the high-scoring compounds was The zebrafish were placed, individually, into docked into this site using the rDock software28, wells in round, 48-well plates (10 mm diameter, and the top-ranked docking pose was extracted. 300 μl of 0.3× Danieau’s solution) after 8 dpf. Three hundred microliters of the sample solutions (8b, su- Cell-Based Reporter Assay vorexant or imipramine) were added to the 48-well HEK293T (human embryonic kidney 293T) plates, which were then placed in an incubator, cells were transfected with human OX1 or OX2 maintained at 28°C with constant light (255 lx). receptors using the Orexin 1 or Orexin 2 Receptor After 3 h, the fishes were anesthetized with 0.05% Assay Kit (Cayman #600240 or #600250, respec- (v/v) ethylene glycol monophenyl ether (Nacalai tively), according to the manufacturer’s instruc- Tesque, Kyoto, Japan). The zebrafish brains were tion. Cells were grown in Dulbecco’s Modified extracted under a stereoscopic microscope (Nikon Eagle’s Medium (DMEM; Nacalai Tesque, Kyoto, SMZ800) and immediately fixed in RNAlater™ Japan) media supplemented with 10% fetal bovine (Applied Biosystems, Foster City, CA, USA) at 4°C serum (FBS; Sigma-Aldrich, St. Louis, MO, USA) for 1 day. The dissected brains were then stored at at 37°C with 5% CO2. Cells were seeded into the −80°C until RNA extraction took place. The total 96-well plates at a density of 50,000 cells/well and RNA was extracted from ten brain samples, using cultivated at 37°C for 16 h in the culture medium. a RNeasy Micro kit (Qiagen, Hilden, Germany), The following day, solutions of approximately 50 according to the manufacturer’s instructions. μl, containing DMSO (3%), suvorexant (3 μM) and 8b (8.7 μM), were added to the cultured cells in Microarray Analysis the 96-well plates to obtain a final volume of 150 Microarray experiments were performed by μl, resulting in 1 μM of suvorexant and 2.9 μM of an Agilent-certified microarray service provider 8b. The cultured cells were then incubated for ap- (TaKaRa Bio, Otsu, Shiga, Japan). Quantification

5130 A novel orexin antagonist from a natural plant was discovered using zebrafish behavioural analysis and quality analysis of RNA were confirmed used to the Gene Expression Omnibus (GEO) database ing an Agilent 2100 Bioanalyzer (Agilent, Santa (GEO accession number GSE129812) run by the Na- Clara, CA, USA). tional Center for Biotechnology Information (NCBI). Microarray experiments were carried out on three independent trials. Total RNA (100 ng) was converted into labelled cRNA using the Low Input Results Quick Amp Labelling Kit (Agilent, Santa Clara, CA, USA). The Cy3-labelled cRNA was hybridized onto Behavioural Profiling of Zebrafish Agilent Zebrafish Oligo DNA Microarrays ver.3.0 ac- Changes in zebrafish behaviour were anal- cording to the manufacturer’s protocol. Microarrays ysed for seven natural compounds, obtained from were scanned using the Agilent SureScan Microar- plants, and were compared to those of known ray Scanner (G2600D) and analysed using the Agi- psychoactive drugs. The behavioural activities of lent Feature Extraction Software 12.0.3.1. Data were the larvae were calculated, based on the Z-score, normalized using the Agilent GeneSpring 14.9.1 soft- as described in Materials and Methods, and ex- ware package. Identification of gene sets differentially amined for 36 behavioural endpoints. One of the expressed in microarray analysis was carried out by compounds, 8b, significantly affected zebrafish one-way ANOVA (p < 0.05). The data were deposit- behaviour. As shown in Figure 2, 8b significantly

Figure 2. Natural compound 8b-(4'-Hydroxytigloyloxy) costunolide affected zebrafish behaviour. The behaviour of zebrafish treated with 8b was classified into three groups based on their mobility. Total distance moved (DMT), in zone frequency (IZF), distance to zone (DTZ), and turn angle (TA) at L1-time, D-time, and L2-time periods are shown for each mobility classification. The data are represented as mean with the standard error of mean. * < 0.05, ** < 0.01 vs. control by Dunnett’s multiple comparisons test.

5131 M. Yamanaka, H. Iwata, K. Masuda, M. Araki, Y. Okuno, M. Okamura, J. Koiwa, T. Tanaka decreased the distance moved total (DMT) scores in receptor antagonist, suggesting that 8b has a at high and low mobility during the dark time mechanism similar to that of suvorexant. In ad- period (D-time), at a concentration of 100 μM. dition, the next best fit with 8b, in behavioural The zone in the centre of the well (DTZ) scores profile, was that of imipramine, indicating that 8b also significantly decreased in the presence of 8b might act on multiple targets. during the D-time. Hierarchical clustering analyses were performed, to test the functional sim- In Silico Assay of Orexin Antagonists ilarity of 8b to the psychoactive drugs, using the The antagonist binding sites were compared, quantitative data matrix from the 36 behavioural between human and zebrafish OX receptors, by endpoints (Z-score). As shown in Figure 3, hier- analysing their amino acid sequences. Based on archical clustering categorized the psychoactive the crystal structures of the OX1 and OX2 recep- drugs according to their mechanism of action. tors, 21 amino acids located within 4 Å from the Thus, psychoactive drugs with a similar mecha- bound suvorexant were defined as pocket-forming nism of action had a similar behavioural profile residues (Figure 4-a, b). Multiple alignment of the in this assay. Clustering analysis showed that the three amino acid sequences of human and zebraf- behavioural profiles by 8b were most similar to ish OX receptors using MAFFT (Multiple Align- those treated with suvorexant, a well-known orex- ment using Fast Fourier Transform)29, showed

Figure 3. Hierarchical clustering of the behaviour of zebrafish treated with psychoactive drugs. The Z scores of each behavioural endpoint were subjected to hierarchical clustering using Pearson correlation with single linkage. Colours shown in the top of hierarchical clustering tree indicate drug class (green, orexin receptor antagonists; light blue, antidepressants; purple, anxiolytics; yellow green, melatonin receptor agonists; yellow, antipsychotics).

5132 A novel orexin antagonist from a natural plant was discovered using zebrafish behavioural analysis

A B

Figure 4. Comparison of the antagonist binding site between human and zebrafish OX receptors.A -B, Crystal structure of the (A) OX1 (PDBID: 4ZJ8) and (B) OX2 (PDBID: 4S0V) receptors (grey cartoon) bound to suvorexant (orange sticks). For each structure, 19 residues located within 4 Å from suvorexant are depicted with cyan sticks. C, Binding pose between the OX2 receptor and 8b-(4'-Hydroxytigloyloxy) costunolide. Side view (left) and top view (right) from the extracellular side of the OX2 receptor. 8b-(4'-Hydroxytigloyloxy) costunolide shows orange sticks. Yellow broken lines show hydrogen bonds between 8b and Pro131, Asn324, and His350 of the OX2 receptor. These figures were generated using the Pymol software35. D, Alignment of amino acid residues forming the antagonist binding site in the human and zebrafish OX receptors. Of the 21 residues defined as binding sites in OX1 or OX2 receptors, 17 are conserved among these three receptors.

5133 M. Yamanaka, H. Iwata, K. Masuda, M. Araki, Y. Okuno, M. Okamura, J. Koiwa, T. Tanaka

Figure 5. In vitro assay suggests 8b-(4'-Hydroxytigloyloxy) costunolide as a dual orexin receptor antagonist. HEK293T (human embryonic kidney 293T) cells were seeded into 12-well plates (50,000 cells/well) and treated with orexin A (50 nM), suvorexant (0.75 μM) plus orexin A (50 nM) or 8b 2.2 μM) plus orexin A (50 nM). Cell supernatants were collected after 24 h and activity of SEAP in culture media was examined. Error bars represent the standard error of mean SE. * < 0.05, ** < 0.01 vs. control by the Student’s t-test. that 17 out of the 21 amino acids were conserved ray analysis. The brain tissues of larvae at 8 dpf among the three receptors (Figure 4-c), suggest- were treated with 8b, imipramine, or suvorexant ing similarity in the shape of the binding pockets. for 3 h. Through this analysis, 377 genes were Among the natural compounds used for be- identified using 10 mM of 8b, 1328 genes were havioural analysis, 8b had the highest ranking, in silico. identified using 100 μM of 8b, 50 genes were Figure 4-d shows binding pose of the OX2 re- identified using 1 μM of suvorexant, and 254 ceptor with 8b. 8b formed hydrogen bonds with genes were identified using 1 μM of imipramine Pro131, Asn324, and His 350 of the OX2 recep- that were differentially up- or down-regulated tor. These amino acids were conserved between compared with the DMSO-treated larvae (p < human and zebrafish OX2 receptors (Figure 4-c). 0.05, fold change ≥ 2). As shown in Figure 6, hier- These analyses resulted in the identification of 8b archical analysis using microarray data indicated as a candidate for an OX1 and OX2 receptor an- that 10 μM of 8b was most closely clustered with tagonist. suvorexant, and second most closely clustered with imipramine. In vitro Reporter Assay with 8b- (4’-Hydroxytigloyloxy) Costunolide Integration of Behavioural Profiling An in vitro reporter assay was conducted to and Microarray Analysis determine whether 8b has an antagonistic activity To determine whether changes in the behaviour towards OX1 and OX2 receptors. The antagonis- pattern are correlated with changes in gene ex- tic function of 8b towards orexin A was evaluated pression, the correlation coefficient between the on a recombinant expressed OX1 or OX2 receptor log2 fold change of gene expression and Z-score in HEK293T. Results of the assay showed that 8b of the behaviour profile, when treated with 10 μM significantly reduced the reporter expression of of 8b was calculated. We then calculated the per- both OX1 and OX2 receptors, while suvorexant centage of the number of gene symbols (NGS) reduced the reporter expression to a much higher whose absolute values of correlation coefficient degree (Figure 5). This result suggests that 8b has were above 0.85, to determine which behavioural the capacity to bind to OX1 and OX2 receptors. parameters are correlated with changes in gene expression. The total number of gene symbols Microarray Expression Analysis was 1340. We used the following formula: Using 8b-(4’-Hydroxytigloyloxy) Costunolide To reveal the molecular mechanism of 8b, the (NGS whose absolute values of 100 × gene expression change was analysed by microar- correlation coefficient were above 0.85)/1340

5134 A novel orexin antagonist from a natural plant was discovered using zebrafish behavioural analysis

Figure 6. Hierarchical clustering of gene expression profile shows that 10 μM of 8b-(4'-Hydroxytigloyloxy) costunolide was closely clustered with suvorexant. Hierarchical clustering was performed using Ward's method and Chebyshev distance metric.

We found that only two parameters had calcu- the orexin receptors, was identified by phenotyp- lated values above 15%: total distance of move- ic screening, using zebrafish behavioural analy- ment with low mobility during the dark time sis. The behavioural analysis performed in this (L-DMT D) and turn angle with medium mo- study classified the central nervous system drugs bility during the dark time (M-TA D). The cas- based on their mechanism of action. Zebrafish es of increase in both the log2 fold-change and treated with 8b showed behaviour profiles simi- the Z-score, or of decrease in both the log2 fold- lar to those of zebrafish treated with suvorexant, change and the Z-score, were referred to as a a known orexin antagonist. In addition, the next “correlation.” Conversely, if the log2 fold change best fit with 8b, in behavioural profile, was that of or Z-score showed opposite changes, it was re- imipramine, indicating that 8b might act on mul- ferred to as an “inverse correlation.” A clustered tiple targets. 8b is contained in herbal medicines, correlation analysis showed that L-DMT and from Eupatorium lindleyanum and Inula heleni- M-TA had significant correlation or inverse cor- um. Extracts from these plants have been used relation to gene expression changes (Figure 7). As in traditional medicines for a long time and are shown in Figure 2, 8b significantly decreased the thought to have positive effects on the respiratory DMT scores at low mobility, during the D-time. and nervous systems. Therefore, a list of the top 30 probes, showing the Through an in silico assay, 8b was confirmed most significant positive or inverse correlations to be a novel, high-potency antagonist for orexin in L-DMT, was compiled to construct functional receptors. In silico assaying is a powerful tool for networks, using ingenuity pathway analysis (IPA, high-throughput screening in the target-based ap- Ingenuity Systems, CA, USA). Through this, it proach30. However, limitations should be consid- was found that NF-κB is a significant hub mol- ered when interpreting the organism’s activity as ecule involved in the gene networks of L-DMT a whole. We believe that integration of zebrafish treated with 10 μM of 8b (Figure 8). phenotypic screening, combined with in silico assaying, shows considerable potential, in drug discovery, to utilize the advantages of both methods. Discussion In vitro screening is a well-established tech- nique used in drug discovery. The demonstration A novel, natural compound, capable of mod- of antagonistic function by 8b to OX1 and OX2 ulating the sleep/wake behaviour, probably via receptors was supported by the in vitro assay. In

5135 M. Yamanaka, H. Iwata, K. Masuda, M. Araki, Y. Okuno, M. Okamura, J. Koiwa, T. Tanaka

Figure 7. A clustered correlation analysis identifying the behavioural analysis parameters correlated to gene expression changes. Each square of the clustergram represents the correlation coefficient between log2 fold change of gene expression and Z score of behaviour profile.

addition, the gene expression profiles of brain tis- that total distance of movement with low mobility sue, from 10 μM 8b-treated larvae, showed closest (L-DMT) and turning angle (TA) correlate closely similarity to those treated with suvorexant, and with gene expression changes in brain tissues. Since second closest similarity with those treated with the exposure to central nervous system depressants imipramine, the same as that described by their such as triazolam and zolpidem causes the reduc- behaviour profiles. Although the gene expression tion of DMT scores and the increase of TA scores profiles by 100 μM 8b-treated larva did not show in zebrafish14, the effect of 8b may be due to its mo- similarity with those treated with suvorexant, it lecular mechanism partially relative to the effects of may be due to nonselective binding to other re- hypnotic drugs. Transcriptome data from 8b-treated ceptors, at high concentrations. larva were all obtained in a narrow time period, just Zebrafish exhibit various behavioural patterns, prior to changing from L1-time to D-time, and not at which appear to be equivalent to human behaviour, other times. To understand the relationship between at a functional level. Moreover, drug screening us- the change in gene expression caused by 8b and the ing zebrafish larvae allows high-throughput analysis behaviour pattern shown during daily light cycle, it of behavioural phenotypes, so that large sets of data would be necessary to collect data sets at multiple can be obtained. Data obtained using transcriptome time points. and behavioural analysis were used to identify the Based on our analysis, NF-κB, a hub molecule, molecular mechanism underlying specific types of was proposed to be involved in the decrease of behaviour phenotypes in zebrafish. Application of DMT scores in the presence of 10 μM of 8b. It has this approach to 10 μM 8b-treated larva showed been reported that inflammatory cytokines, such

5136 A novel orexin antagonist from a natural plant was discovered using zebrafish behavioural analysis

Figure 8. Pathway analysis related to total distance of movement with low mobility (L-DMT) shows that 8b-(4'-Hydroxyti- gloyloxy) costunolide may be involved in inflammatory immune responses. Direct interaction is in solid line, whereas indirect interaction is indicated by a dotted line. The relationship related to NF-κB is highlighted in light blue. The genes in green were down-regulated and the genes in red were up-regulated. The intensity of the node colour correlates with the degree of gene upregulation. IPA network legend is on the right side. as tumour necrosis factor (TNF) and interleukin pression/motivation and learning. To expand this (IL)-1β, can facilitate sleep via the NF-κB path- strategy to other behavioural assays, there are way31-33. In fact, the activity of orexin-expressing several issues that need to be addressed, such as neurones is inhibited by TNF34. extrapolation of zebrafish data to humans. We ex- pect the strategy described here to be a stepping stone in the development of new compounds for treatment of conditions of the nervous system. Conclusions

From this study, zebrafish phenotypic screening, Author Contributions which can be combined with an in silico assay and MY conceived the study, analysed the data, and wrote the gene expression profiling, provides a feasible strat- manuscript. KM, HI, MA, and YO performed the in silico egy for discovering and characterizing novel thera- assay. MO and JK performed the experiments. TT con- peutic compounds, including natural products. ceived the study and wrote the manuscript. The strategy described here could be applied to identify new compounds and their molecular Conflict of Interests mechanism, not only for sleep-wake regulation but also for other types of behaviours, such as de- The author(s) declare that they have no conflict of interests.

5137 M. Yamanaka, H. Iwata, K. Masuda, M. Araki, Y. Okuno, M. Okamura, J. Koiwa, T. Tanaka

References 15) Bennett T, Bray D, Neville MW. Suvorexant, a dual orexin receptor antagonist for the management of 1) Swinney DC, Anthony J. How were new medicines insomnia. P T 2014; 39: 264-266. discovered? Nat Rev Drug Discov 2011; 10: 507-519. 16) Neubauer DN. Almorexant, a dual orexin receptor 2) Bruni G, Lakhani P, Kokel D. Discovering novel antagonist for the treatment of insomnia. Curr neuroactive drugs through high-throughput be- Opin Investig Drugs 2010; 11: 101-110. havior-based chemical screening in the zebrafish. 17) Renzulli C, Nash M, Wright M, Thomas S, Zamuner Front Pharmacol 2014; 5: 153. S, Pellegatti M, Bettica P, Boyle G. Disposition and 3) Dubińska-Magiera M, Daczewska M, Lewicka A, Migo- metabolism of [14C] SB-649868, an orexin 1 and cka-Patrzałek M, Niedbalska-Tarnowska J, Jagla K. 2 receptor antagonist, in humans. Drug Metab Zebrafish: a model for the study of toxicants Dispos 2011; 39: 215-227. affecting muscle development and function. Int J 18) Smart D, Sabido-David C, Brough SJ, Jewitt F, Johns Mol Sci 2016; 17: 1941. A, Porter RA, Jerman JC. SB-334867-A: the first 4) Khan KM, Collier AD, Meshalkina DA, Kysil EV, selective orexin-1 receptor antagonist. Br J Phar- Khatsko SL, Kolesnikova T, Morzherin YY, Warnick JE, macol 2001; 132: 1179-1182. Kalueff AV, Echevarria D. Zebrafish models in neu- 19) Langmead CJ, Jerman JC, Brough SJ, Scott C, Porter ropsychopharmacology and CNS drug discovery. RA, Herdon HJ. Characterisation of the binding of Br J Pharmacol 2017; 174: 1925-1944. [3H]-SB-674042, a novel nonpeptide antagonist, 5) Baraban SC, Dinday MT, Hortopan GA. Drug screen- to the human orexin-1 receptor. Br J Pharmacol ing in Scn1a zebrafish mutant identifies clemizole 2004; 141: 340-346. as a potential Dravet syndrome treatment. Nat 20) Gozzi A, Turrini G, Piccoli L, Massagrande M, Aman- Commun 2013; 4: 2410. tini D, Antolini M, Martinelli P, Cesari N, Montanari Beuckmann CT, Yanagisawa M. 6) Orexins: from neuro- D, Tessari M, Corsi M, Bifone A. Functional mag- peptides to energy homeostasis and sleep/wake netic resonance imaging reveals different neural regulation. J Mol Med 2002; 80: 329-342. substrates for the effects of orexin-1 and orexin-2 7) Chieffi S, Carotenuto M, Monda V, Valenzano A, Vil- receptor antagonists. PLoS One 2011; 6: e16406. lano I, Precenzano F, Tafuri D, Salerno M, Filippi N, 21) Hirose M, Egashira S, Goto Y, Hashihayata T, Ohtake Mollica MP, Lacono D, Nigro E, Monda M, Messina N, Iwaasa H, Hata M, Fukami T, Kanatani A, Yamada G, Messina A. Orexin system: the key for a healthy K. N-acyl 6, 7-dimethoxy-1, 2, 3, 4-tetrahydroiso- life. Front Physiol 2017; 8: 357. quinoline: the first orexin-2 receptor selective 8) Nixon JP, Mavanji V, Butterick TA, Billington CJ, non-peptidic antagonist. Bioorg Med Chem Lett Kotz CM, Teske JA. Sleep disorders, obesity, and 2003; 13: 4497-4499. aging: the role of orexin. Ageing Res Rev 2015; 22) Roecker AJ, Reger TS, Mattern MC, Mercer SP, Berg- 20: 63-73. man JM, Schreier JD, Cube RV, Cox CD, Li D, Lemaire 9) Kornum BR, Faraco J, Mignot E. Narcolepsy with W, Bruno JG, Harrell CM, Garson SL, Gotter AL, hypocretin/orexin deficiency, infections and auto- Fox SV, Stevens J, Tannenbaum PL, Prueksaritanont T, immunity of the brain. Curr Opin Neurobiol 2011; Cabalu TD, Cui D, Stellabott J, Hartman GD, Young 21: 897-903. SD, Winrow CJ, Renger JJ, Coleman PJ. Discovery of 10) Barateau L, Dauvilliers Y. Recent advances in MK-3697: a selective orexin 2 receptor antagonist treatment for narcolepsy. Ther Adv Neurol Disord (2-SORA) for the treatment of insomnia. Bioorg 2019; 12: 1-12. Med Chem Lett 2014; 24: 4884-4890. Wolber G, Dornhofer AA, Langer T 11) Citrome L. Suvorexant for insomnia: a systematic 23) . Efficient overlay review of the efficacy and safety profile for this of small organic molecules using 3D pharmacoph- newly approved hypnotic - what is the number ores. J Comput Aided Mol Des 2006; 20: 773-788. needed to treat, number needed to harm and 24) Wolber G, Langer T. LigandScout: 3-D pharmaco- likelihood to be helped or harmed? Int. J Clin phores derived from protein-bound ligands and Pract 2014; 68: 1429-1441. their use as virtual screening filters. J Chem Inf 12) Zhang B, Shimada Y, Kuroyanagi J, Nishimura Y, Model 2005; 45: 160-169. Umemoto N, Nomoto T, Shintou T, Miyazaki T, Tana- 25) Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat ka T. Zebrafish xenotransplantation model for TN, Weissig H, Shindyalov IN, Bourne PE. The Protein cancer stem-like cell study and high-throughput Data Bank. Nucleic Acids Res 2000; 28: 235-242. screening of inhibitors. Tumour Biol 2014; 35: 26) Yin J, Babaoglu K, Brautigam CA, Clark L, Shao Z, 11861-11869. Scheuermann TH, Harrell CM, Gotter AL, Roecker 13) Westerfield MA. A guide for the laboratory use AJ, Winrow CJ, Renger JJ, Coleman PJ, Rosenbaum of zebrafish (Danio rerio). University of Oregon DM. Structure and ligand-binding mechanism of Press, 2007. the human OX1 and OX2 orexin receptors. Nat 14) Nishimura Y, Okabe S, Sasagawa S, Murakami S, Ash- Struct Mol Biol 2016; 23: 293-299. ikawa Y, Yuge M, Kawaguchi K, Kawase R, Tanaka T. 27) Yin J, Mobarec JC, Kolb P, Rosenbaum DM Yin. Crys- Pharmacological profiling of zebrafish behavior tal structure of the human OX2 orexin receptor using chemical and genetic classification of sleep- bound to the insomnia drug suvorexant. Nature wake modifiers. Front Pharmacol 2015; 6: 257. 2015; 519: 247-250.

5138 A novel orexin antagonist from a natural plant was discovered using zebrafish behavioural analysis

28) Ruiz-Carmona S, Alvarez-Garcia D, Foloppe N, Gar- 31) Bryant PA, Trinder J, Curtis N. Sick and tired: Does mendia-Doval AB, Juhos S, Schmidtke P, Barril X, sleep have a vital role in the immune system? Nat Hubbard RE, Morley SD. rDock: a fast, versatile Rev Immunol 2004; 4: 457-467. and open source program for docking ligands to 32) Krueger JM, Rector DM, Churchill L. Sleep and proteins and nucleic acids. PLoS Comput Biol cytokines. Sleep Med Clin 2007; 2: 161-169. 2014; 10: e1003571. 33) Opp MR. Cytokines and sleep. Sleep Med Rev Katoh K, Standley DM. 29) MAFFT multiple sequence 2005; 9: 355-364. alignment software version 7: improvements in Clark IA, Vissel B performance and usability. Mol Biol Evol 2013; 34) . Inflammation-sleep interface in 30: 772-780. brain disease: TNF, insulin, orexin. J Neuroinflam 2014; 11: 51. 30) Rognan D. The impact of in silico screening in the discovery of novel and safer drug candidates. 35) DeLano WL. The PyMOL molecular graphics sys- Pharmacol Ther 2017; 175: 47-66. tem. DeLano Scientific, 2012.

5139